Read Aloud.
Reading a chapter out loud is the surest way to hear the sentence that runs out of breath or the line of dialogue nobody would actually say, and it is the editing pass most writers skip because a chapter takes the better part of an hour. Read Aloud does the reading with a neural voice that runs entirely on your machine, keeps the sentence being spoken underlined on the page, and in cast mode reads each character's dialogue in a voice you assign.
The shuttle was late, which meant the recycled air had a full ten minutes to find the gap in Mara's collar. She counted the lamps along the north pier instead of counting the minutes.
"You could wait inside, Captain," Hale said from the doorway of the customs office, where it was warm and he had no intention of leaving.
"I could," Mara said, and did not.
The docking horn came first, then the running lights, then the slow white shape of the Harmony's shuttle sliding out of the dark like something remembered. "That will be the Envoy," Hale said.
Reading and navigation.
Voice casting.
When you switch Cast on, quoted speech is read in the voice you assigned to the character who says it, while the dialogue tags and the action around them stay with the narrator, the way an audiobook handles a scene. To work out who is speaking, Draftline tracks quote spans and reads the tag in either order, the action beat before a quote, the nearest mentioned name, and the back-and-forth of two speakers taking turns, and a first pass learns each character's pronoun from the narration so that a nearby name of the wrong gender cannot take a line tagged "she said."
The cast list comes from the Characters Codex. You assign voices one row at a time or press Auto-cast from Characters, which gives every speaker a voice matched to their pronoun and never overrides a choice you made by hand. Assignments are saved inside the book file. A line with no tag, no nearby name, and no turn-taking pattern to follow goes to an explicit Unknown speaker entry and is read by the narrator rather than guessed.
Eight English voices.
Read Aloud runs Kokoro, an open neural voice model, on a native runtime inside Draftline, and generates 24 kHz audio on your own hardware. These are voices with breath and cadence in them rather than the flat system voice most operating systems ship, and you can preview any of them while playback is paused.
Installation and privacy.
Read Aloud is off in a fresh install. When you enable it, Draftline offers a one-time download of the voice model, between 370 and 385 MiB depending on your platform, from pinned sources where every file has a fixed size and SHA-256 hash that is checked before the model is allowed to play. After that the network is not involved. Synthesis runs in a local service bound only to your own machine, on an address that changes every launch, and neither manuscript text nor audio ever leaves the computer. Where the native engine cannot run, Draftline says so; there is no fallback to a browser engine or to a remote service.
What cast mode reads as narration.
Cast attribution works on dialogue set in double quotation marks, which is what the great majority of published fiction uses. If your manuscript sets dialogue with em dashes, or with single quotes throughout, those lines are read in the narrator's voice rather than attributed, and everything else on this page works the same.
